“Our teams are losing time” is the most common symptom, and the most frequently misdiagnosed. Depending on what sits behind it, the answer is a conversational assistant or an automation — and confusing the two is expensive.
Two tools, two natures
An assistant answers a question asked by a human. It is triggered by a conversation, it produces text, and its output is read before it is used. An automation performs a task without being asked. It is triggered by an event — an email arriving, an order confirmed, a time of day — and its output goes straight into another system.
The test that settles it almost every time
Ask one question: is the task triggered by someone waiting for an answer, or by an event that nobody is waiting on? If a human is waiting, it is an assistant. If nobody is waiting, it is an automation.
The costs are not alike
An automation costs to build and then almost nothing: it runs, you watch it. An assistant costs on every conversation, because a model is called each time, and it demands continuous work on the knowledge base. For the same need, automation is often the cheaper option to run — when it is possible.
The classic trap
The trap is putting an assistant in front of a process that has never been written down. The tool then becomes the only place the rule exists, and nobody can check it. Write the process first; you will often find that half the need is solved by a simple automation, with no model at all.
Where to start
List your ten most repetitive tasks and mark each one “somebody is waiting” or “nobody is waiting”. You will have your roadmap, and the budget that goes with it.